Hyperspectral Imaging Applications Research Articles

Rapid and non-destructive identification of Panax ginseng origins using hyperspectral imaging, visible light imaging, and X-ray imaging combined with multi-source data fusion strategies

The geographical origin of Panax ginseng significantly influences its nutritional value and chemical composition, which in turn affects its market price. Traditional methods for analyzing these differences are often time-consuming and require substantial quantities of reagents, rendering them inefficient. Therefore, hyperspectral imaging (HSI) in conjunction with X-ray technology were used for the swift and non-destructive traceability of Panax ginseng origin. Initially, outlier samples were effectively rejected by employing a combined isolated forest algorithm and density peak clustering (DPC) algorithm. Subsequently, random forest (RF) and support vector machine (SVM) classification models were constructed using hyperspectral spectral data. These models were further optimized through the application of 72 preprocessing methods and their combinations. Additionally, to enhance the model’s performance, four variable screening algorithms were employed: SelectKBest, genetic algorithm (GA), least absolute shrinkage and selection operator (LASSO), and permutation feature importance (PFI). The optimized model, utilizing second derivative, auto scaling, permutation feature importance, and support vector machine (2nd Der-AS-PFI-SVM), achieved a prediction accuracy of 93.4 %, a Kappa value of 0.876, a Brier score of 0.030, an F1 score of 0.932, and an AUC of 0.994 on an independent prediction set. Moreover, the image data (including color information and texture information) extracted from color and X-ray images were used to construct classification models and evaluate their performance. Among them, the SVM model constructed using texture information from X -ray images performed the best, and it achieved a prediction accuracy of 63.0 % on the validation set, with a Brier score of 0.181, an F1 score of 0.518, and an AUC of 0.553. By implementing mid-level fusion and high-level data fusion based on the Stacking strategy, it was found that the model employing a high-level fusion of hyperspectral spectral information and X-ray images texture information significantly outperformed the model using only hyperspectral spectral information. This advanced model attained a prediction accuracy of 95.2 %, a Kappa value of 0.912, a Brier score of 0.027, an F1 score of 0.952, and an AUC of 0.997 on the independent prediction set. In summary, this study not only provides a novel technical path for fast and nondestructive traceability of Panax ginseng origin, but also demonstrates the great potential of the combined application of HSI and X-ray technology in the field of traceability of both medicinal and food products.

Probabilistic Mixture Model-Based Spectral Unmixing

Spectral unmixing attempts to decompose a spectral ensemble into the constituent pure spectral signatures (called endmembers) along with the proportion of each endmember. This is essential for techniques like hyperspectral imaging (HSI) used in environment monitoring, geological exploration, etc. Several spectral unmixing approaches have been proposed, many of which are connected to hyperspectral imaging. However, most extant approaches assume highly diverse collections of mixtures and extremely low-loss spectroscopic measurements. Additionally, current non-Bayesian frameworks do not incorporate the uncertainty inherent in unmixing. We propose a probabilistic inference algorithm that explicitly incorporates noise and uncertainty, enabling us to unmix endmembers in collections of mixtures with limited diversity. We use a Bayesian mixture model to jointly extract endmember spectra and mixing parameters while explicitly modeling observation noise and the resulting inference uncertainties. We obtain approximate distributions over endmember coordinates for each set of observed spectra while remaining robust to inference biases from the lack of pure observations and the presence of non-isotropic Gaussian noise. As a direct impact of our methodology, access to reliable uncertainties on the unmixing solutions would enable robust solutions to noise, as well as informed decision-making for HSI applications and other unmixing problems.

Critical Evaluation of Hyperspectral Imaging Technology for Detection and Quantification of Microplastics in Soil

In this study, we critically evaluated the performance of an emerging technology, hyperspectral imaging (HSI), for detecting microplastics (MPs) in soil. We examined the technology's robustness against varying environmental conditions in five groups of experiments. Our findings show that near-infrared (NIR) hyperspectral imaging (HSI) effectively detects microplastics (MPs) in soil, though detection efficacy is influenced by factors such as MP concentration, color, and soil moisture. We found a generally linear relationship between the levels of MPs in various soils and their spectral responses in the NIR HSI imaging spectrum. However, effectiveness is reduced for certain MPs, like polyethylene, in kaolinite clay. Furthermore, we showed that soil moisture considerably influenced the detection of MPs, leading to nonlinearities in quantification and adding complexities to spectral analysis. The varied responses of MPs of different sizes and colors to NIR HSI present further challenges in detection and quantification. The research suggests pre-grouping of MPs based on size before analysis and proposes further investigation into the interaction between soil moisture and MP detectability to enhance HSI's application in MP monitoring and quantification. To our knowledge, this study is the first to comprehensively evaluate this technology for detecting and quantifying microplastics. Environmental ImplicationMicroplastics (MPs), increasingly recognized as hazardous materials, pose significant ecological and health risks due to their persistence and potential for bioaccumulation. Our research highlights the significant impact of environmental factors on the detection of MPs in soil using spectrometric methods such as hyperspectral imaging (HSI). This aspect is vital yet often neglected in MP monitoring and detection efforts. By evaluating HSI technology for MP detection in various soil conditions, our study contributes to developing effective monitoring methods. Our findings facilitate better monitoring and managing MP pollution, thus addressing a critical environmental problem and protecting terrestrial ecosystems and human health.

Application of Hyperspectral Imaging for Watermelon Seed Classification Using Deep Learning and Scoring Mechanism

Watermelon seeds are a significant source of nutrition in the diet. To assess the potential of near-infrared hyperspectral imaging technology for swift and nondestructive identification of watermelon seed varieties, near-infrared hyperspectral imaging (NIR-HSI) technology was used. The Savitzky–Golay (SG) smoothing algorithm and standard normal variable (SNV) algorithm were combined to preprocess the extracted spectral data. The successive projections algorithm (SPA) was used to reduce the dimensionality of the spectral data. Subsequently, three deep learning models (LeNet, GoogLeNet, and ResNet) were used to classify 10 common watermelon seeds. SPA was used to reduce the dimensionality of hyperspectral data. In terms of full band, the ResNet model achieved a classification accuracy of 86.77% on the test set. By using characteristic bands, the GoogLeNet model achieved a classification accuracy of 83.85% on the test set. The ensemble fusion model based on a scoring mechanism achieved accuracy rates of 99.56%, 90.88%, and 87.97% on the training, validation, and test sets, respectively. The results indicated that the ensemble fusion model based on a scoring mechanism can enhance accuracy. Combining deep learning with NIR-HSI can effectively distinguish different varieties of watermelon seeds.

Fabrication of a high-quality, small blaze angle grating for visible short-wave infrared hyperspectral cameras

This study outlines the development of a low line density, small blaze angle grating, optimized for a visible to short-wave infrared hyperspectral camera. An analysis of grating specifications was conducted to meet the precise requirements of this application, particularly focusing on the stringent tolerance limits for the blaze angle. A specialized ruling tool adjustment device was designed to adhere to these exacting blaze angle tolerances. The grating groove shape was examined using atomic force microscopy (AFM), and the theoretical diffraction efficiency of the grating was calculated based on these observations. Additionally, laser-based methods were employed to measure the actual diffraction efficiency of the grating, while interferometry was used to assess the grating’s diffraction wavefront. The test results demonstrate our capability to fabricate high-quality gratings with a low line density and small blaze angles that are suitable for advanced hyperspectral imaging applications.

Toward more efficient and effective color quality control for the large‐scale offset printing process

AbstractThis study illustrates at‐line application of hyperspectral imaging in the visible range for quality control of large‐scale offset printing. In particular, the measurement stability of a competitive device is assessed and compared to traditional handheld and desktop spectrophotometers. The performance of the commercially available instruments was assessed based on collected spectra and their corresponding L*, a*, and b* values. The printing process was described by hyperspectral images (in visible range) of selected regions from template color fields acquired at 17 sampling occasions. Spectra constituting hyperspectral images were visualized and evaluated in the space of significant principal components obtained from the principal component analysis. Furthermore, confidence ellipses were constructed for each set of spectra characterizing a specific moment of the printing process. Comparing their mutual locations, shapes, orientations, and sizes enabled effective visualization of process variability and was more comprehensive regarding the classic approach based on information provided by desktop and handheld spectrometers.

Application of hyperspectral imaging for characterization of VOC-induced historical glass corrosion

Hyperspectral imaging (HSI) provides high-resolution recording of the spectral information at every spatial point (pixel) of an object throughout a contiguous range of wavelengths. This paper reports an attempt at exploring the application of HSI in transmittance mode for the identification and quantification of corrosion in transparent historical glass, which is difficult to identify by visual inspection. This was done using model glass samples mimicking historical composition and subjected to artificial ageing in the presence of volatile organic compounds. Hyperspectral images of unaged and aged glass samples were recorded with two cameras covering visible and near infrared (VNIR, 400–1000 nm) and short-wave infrared range (SWIR, 1000–2500 nm) using a custom-made HSI set-up in transmission mode. The HSI data were further processed to classify and visualize corrosion in different ageing environments and ageing periods. The results show that HSI can be employed as a valuable tool to assess glass corrosion at early stages, especially by analysing the SWIR spectral region—which shows the water absorption by the glass matrix due to corrosion.

Application of hyperspectral imaging and colorimetric sensor array coupled with multivariate analysis for quality detection during salted duck eggs processing

AbstractIn order to realize the evaluation of salted duck eggs quality, this study constructed salts‐sensitive colorimetric sensor array (CSA) based on silver nanoprisms (AgNPRs) etching principle and utilized hyperspectral imaging (HSI) system to capture the reflectance spectral of CSA and section of salted duck eggs. To minimize the noise in spectral signals, three preprocessing methods including Savitzky–Golay Smoothing (SG), standard normal variate (SNV) transformation, and multiplicative scatter correction were used. The competitive adaptive reweighted sampling and uninformative variables elimination were used to select the optimal wavelengths, and then partial least squares regression was used to predict moisture, lipid, and salt content (SC). Results demonstrated coefficients of determination of 0.8200, 0.8969, 0.8709, and 0.9239 for protein moisture, yolk moisture, yolk lipid, and protein SC. Visualization distribution maps successfully display the distribution of moisture and lipid content during salting periods. The results indicated that the CSA based on AgNPRs etching principle can be an effective tool in quantitative measurement of salt and HSI combining CSA can be applied to evaluate the quality of salted duck eggs comprehensively. The study offers a rapid detection method for the production and processing of salted duck eggs in the industry.Practical applicationA novel salts‐sensitive CSA was constructed to evaluate quality of salted duck eggs based on AgNPRs etching. This study designed and developed a portable sample collection device using 3D printing technology to acquire the information of samples and CSA at the same time. This study combined with HSI system and colorimetric sensor technology to evaluate the quality of salted duck eggs during the salting process.

Hyperspectral imaging-based erythema classification in atopic dermatitis.

Among the characteristics that appear in the epidermis of the skin, erythema is primarily evaluated through qualitative scales, such as visual assessment (VA). However, VA is not ideal because it relies on the experience and skill of dermatologists. In this study, we propose a new evaluation method based on hyperspectral imaging (HSI) to improve the accuracy of erythema diagnosis in clinical settings and investigate the applicability of HSI to skin evaluation. For this study, 23 subjects diagnosed with atopic dermatitis were recruited. The inside of the right arm is selected as the target area and photographed using a hyperspectral camera (HS). Subsequently, based on the erythema severity visually assessed by a dermatologist, the severity classification performance of the RGB and HS images is compared. Erythema severity is classified as high when using (i) all reflectances of the entire HSI band and (ii) a combination of color features (R of RGB, a* of CIEL*a*b*) and five selected bands through band selection. However, as the number of features increases, the amount of calculation increases and becomes inefficient; therefore, (ii), which uses only seven features, is considered to perform classification more efficiently than (i), which uses 150 features. In conclusion, we demonstrate that HSI can be applied to erythema severity classification, which can further increase the accuracy and reliability of diagnosis when combined with other features observed in erythema. Additionally, the scope of its application can be expanded to various studies related to skin pigmentation.

A Systematized Review on the Applications of Hyperspectral Imaging for Quality Control of Potatoes

AbstractThe application of hyperspectral imaging (HSI) has gained significant importance in the past decade, particulary in the context of food analysis, including potatoes. However, the current literature lacks a comprehensive systematic review of the application of this technique in potato cultivation. Therefore, the aim of this work was to conduct a systematized review by analysing the most relevant compounds, diseases and stress factors in potatoes using hyperspectral imaging. For this purpose, scientific studies were retrieved through a systematic keyword search in Web of Science and Scopus databases. Studies were only included in the review if they provided at least one set of quantitative data. As a result, a total of 52 unique studies were included in the review. Eligible studies were assigned an in-house developed quality scale identifying them as high, medium or low risk. In most cases the studies were rated as low risk. Finally, a comprehensive overview of the HSI applications in potatoes was performed. It has been observed that most of the selected studies obtained better results using linear methods. In addition, a meta-analysis of studies based on regression and classification was attempted but was not possible as not enough studies were found for a specific variable.

Non-destructive assessment of soluble solids content in kiwifruit using hyperspectral imaging coupled with feature engineering.

The maturity of kiwifruit is widely gauged by its soluble solids content (SSC), with accurate assessment being essential to guarantee the fruit's quality. Hyperspectral imaging offers a non-destructive alternative to traditional destructive methods for SSC evaluation, though its efficacy is often hindered by the redundancy and external disturbances of spectral images. This study aims to enhance the accuracy of SSC predictions by employing feature engineering to meticulously select optimal spectral features and mitigate disturbance effects. We conducted a comprehensive investigation of four spectral pre-processing and nine spectral feature selection methods, as components of feature engineering, to determine their influence on the performance of a linear regression model based on ordinary least squares (OLS). Additionally, the stacking generalization technique was employed to amalgamate the strengths of the two most effective models derived from feature engineering. Our findings demonstrate a considerable improvement in SSC prediction accuracy post feature engineering. The most effective model, when considering both feature engineering and stacking generalization, achieved an of 0.721, a of 0.046, and an of 1.394 in the prediction set. The study confirms that feature engineering, especially the careful selection of spectral features, and the stacking generalization technique are instrumental in bolstering SSC prediction in kiwifruit. This advancement enhances the application of hyperspectral imaging for quality assessment, offering benefits that extend across the agricultural industry.

Terajet-assisted time-domain super-resolution imaging

Large-scale applications of real-time terahertz (THz) imaging are often limited by the spatial resolution of a THz microscope. One of the simplest approaches to circumvent this problem in practice and achieve resolution below the diffraction limit consists in placing a dielectric particle with a size comparable to the radiation wavelength in the vicinity of the focal plane of the THz imaging system. Here, we theoretically consider the transmittance of a THz Gaussian beam through a Teflon sphere to further analyze experimental results on THz superresolution imaging. We address the resolution beyond the diffraction limit (the experimentally determined value is 0.38λ at 0.46 THz) through the superlens effect of the terajet generated by the mesoscale sphere owing to propagating and evanescent partial waves in the terajet. Due to the coupling of the dielectric sphere and time-domain THz spectrometer, we can detect a 10 µm thick PTFE film, which is almost transparent (98% transmittance) in the THz frequency range. The correlation between the transmittance spectrum of the dielectric sphere and the quality of THz images of thin, nonconducting films collected at different frequencies has been demonstrated experimentally and considered analytically. Our study is relevant for hyperspectral imaging applications and further development of the sphere-coupled time-domain THz imaging technique.

Application of near-infrared hyperspectral imaging for determination of cheese chemical composition

Cheeses vary in chemical composition as they have different applications according to the technological features; some hard cheeses are consumed fresh while melting is a desirable characteristic for others. The conventional reference methods for determination of cheese composition and quality are accurate, but include physical-chemical analyses that are time-consuming and require use of chemical reagents. The objective of this work was to apply near-infrared hyperspectral imaging (NIR-HSI) as a fast method for prediction and visualization of spatial distribution of chemical composition of different types of cheeses (mozzarela, requeijão do norte, minas meia cura, coalho, prato). Moisture, fat, protein, and free oil were measured, and used as reference for prediction models, while spectral information extracted from images in the range of 928–2524 nm were used as predictors. Partial Least Squares (PLS) regression models were built with full spectra and using only few selected regions with higher importance, which were applied to the images to build chemical maps and visualize the spatial variation of the cheese composition. The regression models were able to predict moisture (R2P = 0.92, RMSEP = 1.1%); fat (R2P = 0.84, RMSEP = 1.5%); protein (R2P = 0.80, RMSEP = 1.8%); and free oil (R2P =0.64, RMSEP = 0.5%) for an external set of samples, while requiring a small number of latent variables (LVs) (between 3 and 4). The chemical maps allowed visualization of the spatial distribution for each attribute, demonstrating that the composition may vary within sample. The study demonstrated the potential of NIR-HSI for the determination of cheese composition and possible application to cheese quality control.

Fertility detection of unincubated chicken eggs by hyperspectral transmission imaging in the Vis-SWNIR region

Detection of infertile eggs prior to incubation can lead to an increase in the hatchability rate and prevent the wastage of billions of non-fertile eggs ended up by failed incubation. In this study, the feasibility of a line-scan hyperspectral imaging system in the visible and short-wavelength near-infrared region was assessed for early detection of non-fertile eggs on day 0 before incubation. A total of 227 white-shell eggs including 131 fertile and 96 infertile eggs were collected from a flock with similar conditions in terms of hen age, feeding, and management. Hyperspectral images of eggs were captured on day 0 before incubation in a transmittance mode of illumination and then the eggs were incubated in a commercial incubator. The edge detection method was used to segment the egg, including both the white and yolk, from the background, and the image spectral information was extracted from the egg region. After applying various pretreatment methods, different classifiers including soft independent modeling of class analogy (SIMCA), linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), and artificial neural networks (ANN) classifiers were utilized to extract the predictive models. Following the acceptable results of SIMCA analysis accomplished by 1st derivative pretreatment (accuracy of 86.67%), the discrimination power plot was used to select the most informative wavebands. The results showed that by using fewer variables in effective wavebands better performance (precision and accuracy of 92.59% and 93.33%, respectively) could be obtained in comparison with the ANN classifier based on the whole spectral data (precision and accuracy of 89.29% and 91.11%, respectively). This study revealed the potential application of hyperspectral transmittance imaging in the Vis-SWNIR region to discern the fertile and infertile eggs before starting the incubation process.

Identification of Skin Lesions by Snapshot Hyperspectral Imaging.

This study pioneers the application of artificial intelligence (AI) and hyperspectral imaging (HSI) in the diagnosis of skin cancer lesions, particularly focusing on Mycosis fungoides (MF) and its differentiation from psoriasis (PsO) and atopic dermatitis (AD). By utilizing a comprehensive dataset of 1659 skin images, including cases of MF, PsO, AD, and normal skin, a novel multi-frame AI algorithm was used for computer-aided diagnosis. The automatic segmentation and classification of skin lesions were further explored using advanced techniques, such as U-Net Attention models and XGBoost algorithms, transforming images from the color space to the spectral domain. The potential of AI and HSI in dermatological diagnostics was underscored, offering a noninvasive, efficient, and accurate alternative to traditional methods. The findings are particularly crucial for early-stage invasive lesion detection in MF, showcasing the model's robust performance in segmenting and classifying lesions and its superior predictive accuracy validated through k-fold cross-validation. The model attained its optimal performance with a k-fold cross-validation value of 7, achieving a sensitivity of 90.72%, a specificity of 96.76%, an F1-score of 90.08%, and an ROC-AUC of 0.9351. This study marks a substantial advancement in dermatological diagnostics, thereby contributing significantly to the early and precise identification of skin malignancies and inflammatory conditions.

Research Review on Quality Detection of Fresh Tea Leaves Based on Spectral Technology.

As the raw material for tea making, the quality of tea leaves directly affects the quality of finished tea. The quality of fresh tea leaves is mainly assessed by manual judgment or physical and chemical testing of the content of internal components. Physical and chemical methods are more mature, and the test results are more accurate and objective, but traditional chemical methods for measuring the biochemical indexes of tea leaves are time-consuming, labor-costly, complicated, and destructive. With the rapid development of imaging and spectroscopic technology, spectroscopic technology as an emerging technology has been widely used in rapid non-destructive testing of the quality and safety of agricultural products. Due to the existence of spectral information with a low signal-to-noise ratio, high information redundancy, and strong autocorrelation, scholars have conducted a series of studies on spectral data preprocessing. The correlation between spectral data and target data is improved by smoothing noise reduction, correction, extraction of feature bands, and so on, to construct a stable, highly accurate estimation or discrimination model with strong generalization ability. There have been more research papers published on spectroscopic techniques to detect the quality of tea fresh leaves. This study summarizes the principles, analytical methods, and applications of Hyperspectral imaging (HSI) in the nondestructive testing of the quality and safety of fresh tea leaves for the purpose of tracking the latest research advances at home and abroad. At the same time, the principles and applications of other spectroscopic techniques including Near-infrared spectroscopy (NIRS), Mid-infrared spectroscopy (MIRS), Raman spectroscopy (RS), and other spectroscopic techniques for non-destructive testing of quality and safety of fresh tea leaves are also briefly introduced. Finally, in terms of technical obstacles and practical applications, the challenges and development trends of spectral analysis technology in the nondestructive assessment of tea leaf quality are examined.

Analysis of cellular response to drugs with a microfluidic single-cell platform based on hyperspectral imaging

BackgroundCellular response to pharmacological action of drugs is significant for drug development. Traditional detection method for cellular response to drugs normally rely on cell proliferation assay and metabolomics examination. In principle, these analytical methods often required cell labeling, invasion analysis, and hours of co-culture with drugs, which are relatively complex and time-consuming. Moreover, these methods can only indicate the drug effectiveness on cell colony rather than single cells. Thus, to meet the requirements of personal precision medicine, the development of drug response analysis on the high resolution of single cell is demanded. ResultsTo provide precise result for drug response on single-cell level, a microfluidic platform coupled with the label-free hyperspectral imaging was developed. With the help of horizontal single-cell trapping sieves, hundreds of single cells were trapped independently in microfluidic channels for the purposes of real-time drug delivery and single-cell hyperspectral image recording. To significantly identify the cellular hyperspectral change after drug stimulation, the differenced single-cell spectrum was proposed. Compared with the deep learning classification method based on hyperspectral images, an optimal performance can be achieved by the classification strategy based on differenced spectra. And the cellular response to different reagents, for example, K+, Epidermal Growth Factor (EGF), and Gefitinib at different concentrations can be accurately characterized by the differenced single-cell spectra analysis. Significance and noveltyThe high-throughput, rapid analysis of cellular response to drugs at the single-cell level can be accurately performed by our platform. After systematically analyzing the materials and the structures of the single-cell microfluidic chip, the optimal single-cell trapping method was proposed to contribute to the further application of hyperspectral imaging on microfluidic single-cell analysis. And the hyperspectral characterization of single-cell with cancer drug stimulation proved the application potential of our method in personal cancer medication.

Unraveling Molecular Composition in Biological Samples-Systematic Evaluation of Statistical Methods for the Analysis of Hyperspectral Raman Data.

Recently, confocal Raman microscopy has gained popularity in biomedical research for studying tissues in healthy and diseased state due to its ability to acquire chemically selective data in a noninvasive approach. However, biological samples, such as brain tissue, are inherently difficult to analyze due to the superposition of molecules in the Raman spectra and low variation of spectral features within the sample. The analysis is further impeded by pathological hallmarks, for example beta-amyloid (Aβ) plaques in Alzheimer's disease, which are often solely characterized by subtle shifts in the respective Raman peaks. To unravel the underlying molecular information, convoluted statistical procedures are inevitable. Unfortunately, such statistical methods are often inadequately described, and most natural scientists lack knowledge of their appropriate use, causing unreproducible results and stagnation in the application of hyperspectral Raman imaging. Therefore, we have set out to provide a comprehensive guide to address these challenges with the example of a complex hyperspectral data set of brain tissue samples with Aβ plaques. Our study encompasses established as well as novel statistical methods, including univariate analysis, principal component analysis, cluster analysis, spectral unmixing, and 2D correlation spectroscopy, and critically compares the outcomes of each analysis. Moreover, we transparently demonstrate the effect of preprocessing decisions like denoising and scaling techniques, providing valuable insights into implications of spectral quality for data evaluation. Thereby, this study provides a comprehensive evaluation of analysis approaches for complex hyperspectral Raman data, laying out a blueprint for elucidating meaningful information from biological samples in chemical imaging.

The Identification of Fritillaria Species Using Hyperspectral Imaging with Enhanced One-Dimensional Convolutional Neural Networks via Attention Mechanism.

Combining deep learning and hyperspectral imaging (HSI) has proven to be an effective approach in the quality control of medicinal and edible plants. Nonetheless, hyperspectral data contains redundant information and highly correlated characteristic bands, which can adversely impact sample identification. To address this issue, we proposed an enhanced one-dimensional convolutional neural network (1DCNN) with an attention mechanism. Given an intermediate feature map, two attention modules are constructed along two separate dimensions, channel and spectral, and then combined to enhance relevant features and to suppress irrelevant ones. Validated by Fritillaria datasets, the results demonstrate that an attention-enhanced 1DCNN model outperforms several machine learning algorithms and shows consistent improvements over a vanilla 1DCNN. Notably under VNIR and SWIR lenses, the model obtained 98.97% and 99.35% for binary classification between Fritillariae Cirrhosae Bulbus (FCB) and other non-FCB species, respectively. Additionally, it still achieved an extraordinary accuracy of 97.64% and 98.39% for eight-category classification among Fritillaria species. This study demonstrated the application of HSI with artificial intelligence can serve as a reliable, efficient, and non-destructive quality control method for authenticating Fritillaria species. Moreover, our findings also illustrated the great potential of the attention mechanism in enhancing the performance of the vanilla 1DCNN method, providing reference for other HSI-related quality controls of plants with medicinal and edible uses.

Identification of geographical origin of semen ziziphi spinosae based on hyperspectral imaging combined with convolutional neural networks

The accurate identification of geographical origin is one of the main challenges in quality assessment of semen ziziphi spinosae. This paper proposed resnet50-based and shuffle-based convolutional neural network (CNN) models (CNN-resnet and CNN-shuffle for hyperspectral image input, as well as CNN-resnet1D and CNN-shuffle1D for spectra input) to discriminate semen ziziphi spinosae from five geographical origins. Specifically, minimum noise fraction combined with mask was adopted to remove background pixels. Savitzky-Golay smoothing, followed by standard normal variate, was applied to denoise in pixel-wise. Mahalanobis distances were calculated based on the average spectra to identify and exclude outliers. Partial least squares discriminant analysis and support vector machine were also adopted to compare with the four CNN networks, particular in model accuracy, computing time, and parameters. Overall, CNN-shuffle was the best, with accuracy (0.902) comparable to CNN-resnet and in less time (0.348 s) and with fewer parameters (5607). Feature wavelengths were selected based on the learned weight coefficients in self-defined layer of CNN-resnet1D and CNN-shuffle1D models. They were related to lipid and protein, of which contents were statistically analyzed. Both lipids and proteins were key parameters, and lipids played a greater role in the identification. Spinosin as a quality marker was measured via standard HPLC method, which was comparably discussed with the hyperspectral imaging (HSI). The potential application of HSI with CNN in quality control of semen ziziphi spinosae within the modern system of mass production was higher compared with HPLC.